Global Financial Interconnectedness and Spillovers Between the G20 Countries

Deutsche Bundesbank Monthly Report September 2020 53

Global financial interconnectedness and spillovers between the G20 countries

Even after the global financial crisis of 2008-09, the international interconnectedness of national financial systems continued to deepen, albeit at a slower pace than before. The extent to which the coronavirus pandemic has influenced this trend can only be assessed in greater detail once a certain amount of time has passed. Despite this, the pandemic has once again emphatically raised the question of the extent to which the deepening interconnectedness between financial systems has changed advanced and emerging market economies’ susceptibility to shocks.

A look at the development of global capital flows and their volatility shows that an abrupt out- flow or reversal of capital flows can pose significant challenges, particularly for countries with less developed financial systems. As demonstrated by the global financial crisis, however, financial crises and the risk of contagion are not problems that affect only emerging market economies. As open economies are interconnected in financial and real economic terms, shocks in one country can also have an impact on other countries, which can in turn have feedback effects on the country in which the shock originated. To an increasing degree, this also applies to the relationships between advanced economies and emerging market economies. The analysis of spillover effects in the equity markets of the G20 countries shows that spillovers from advanced economies to emerging market economies continue to surpass those from emerging market economies to advanced economies. It is also shown that spillover effects can rise sharply and abruptly during periods of stress, such as the present phase triggered by the outbreak of the COVID-19 pandemic. Furthermore, the COVID-19 pandemic has also revealed increased vulnerabilities in individual G20 emerging market economies.

First and foremost, this has presented substantial challenges to national economic policy. Along- side stability-oriented economic policy, the further expansion of local capital markets, the build- up of foreign currency reserves and the implementation of macroprudential measures are appropriate for containing these vulnerabilities. Capital flow measures can also help to ensure financial stability. However, national policy measures alone are not always sufficient to combat the nega- tive repercussions of highly volatile capital flows. In such cases, assistance from the international community, such as financial support from the International Monetary Fund (IMF), can supple- ment national efforts. Deutsche Bundesbank Monthly Report September 2020 54

Development of cross-border netted out and therefore may also have nega- capital flows tive values.1 This change in focus is due, amongst other reasons, to the fact that both Growing capital The international interconnectedness of finan- components of the net figure – gross inflows flows increase cial systems has risen considerably since the from non-resident investors and gross outflows interconnectedness within the 1980s. While the majority of international cap- from resident investors – are generally larger global financial system ital flows still take place between advanced and more volatile than net inflows. They can economies, the share of capital flows attribut- thus be indicative of spillover effects and vul- able to emerging market economies has risen nerabilities that, due to netting, are not neces- to a globally significant level. This development sarily reflected in equivalent changes in net has been driven by advances in information capital flows. In addition, looking at gross fig- and communication technologies, which have ures allows for an analysis of different behav- made it considerably easier to transfer assets ioural patterns between resident and non- across borders. In addition, emerging market resident investors. For example, the economic economies have sought to take advantage of policy implications of strong net capital inflows the benefits of international capital flows by can differ depending on whether these inflows opening up and developing their financial mar- are the result of increased investment from kets, which had previously been closed. abroad or the repatriation of funds by resident investors.2 International Capital flows can have a positive impact on the capital flows recipient country if foreign capital is used to The development of capital flows is shown in Following a can have sharp rise before considerable fund investment and stimulate economic the balance of payments data for the G20 the global finan- benefits … growth. Furthermore, an international disper- countries (see the chart on p. 55).3 In this con- cial crisis, gross capital inflows sion of financial assets allows investors to pur- text, the strongest growth in gross capital flows have since seen chase higher-yielding assets as well as to better worldwide was recorded in the first few years more subdued diversify risks and thereby reduce the total risk of the new millennium. development … of investment. The trend of rising capital flows was inter- … but also However, for emerging market economies in rupted by the global financial crisis. Following harbour risks particular, inflows and outflows of capital can the collapse of Lehman Brothers in September often present major challenges to financial sta- 2008, there was initially a massive slump in bility, especially if the capital flows exhibit a global gross capital inflows. Although these high level of volatility. In relatively underdeveloped institutional environments, sudden stops in inflows or sharp rises in outflows can more 1 “Gross capital inflows” refer to the purchases less sales of domestic assets by non-residents within a specific period. quickly result in financial and currency crises, Accordingly, “gross capital outflows” are the purchases less sometimes at considerable cost to the affected sales of foreign assets by residents. “Net capital inflows” are the difference between gross capital inflows and gross countries. However, very sharp increases in in- capital outflows. See Committee on the Global Financial flows can also pose challenges to macroeco- System (2009). Net capital flows can be interpreted as the financial counterpart to the current account balance and nomic and financial stability. thus provide an opportunity for approximating the impact of non-resident investors on the domestic economy. See Borio and Disyatat (2015). Increasing focus In order to more effectively identify potential 2 See Obstfeld (2014) and Forbes and Warnock (2012). on gross flows vulnerabilities, the analysis of capital flows has 3 Here and in the rest of this article, the focus is on the instead of net countries belonging to the Group of 20 (G20). The Euro- flows in order to changed in recent years. Up until the global pean Union, which is also a G20 member, is disregarded. assess potential financial crisis, analyses had centred mostly on The G20 emerging market economies comprise Argentina, vulnerabilities Brazil, China, India, Indonesia, Mexico, Russia, Saudi Ara- net capital flows; since then, the focus has bia, South Africa and Turkey. The advanced economies in shifted to gross capital flows. Here, it should be the G20 are Australia, Canada, France, Germany, Italy, Japan, South Korea, the United Kingdom and the United noted that the gross flows have already been States. Deutsche Bundesbank Monthly Report September 2020 55

Gross capital inflows

As a percentage of total GDP of the G20 countries

+ 6 Emerging market economies Advanced economies + 5

+ 4

+ 3

+ 2

+ 1

– 1

– 2

– 3

1990 95 00 05 10 15 2019

Source: IMF Balance of Payments. Deutsche Bundesbank

subsequently recovered, they have grown availability of balance of payments data. Never- much more slowly than they had before the theless, in March this year, the inflows to in- global financial crisis. Over time, a growing vestment funds that invest in equity and bond proportion of gross capital inflows have been funds in G20 emerging market economies fell attributable to the emerging market econ- by an unprecedented amount. These capital omies.4 Against the backdrop of adjustments flows are generally considered to be a good to the financial system in the wake of the approximation of portfolio investment.8 In the global financial crisis, the rising attractiveness second quarter of 2020, strong outflows from of the emerging market economies to investors equity funds continued to be observed, while led to comparatively robust capital inflows, inflows to bond funds stabilised somewhat. while the inflows to and outflows from the ad- Compared with other crisis episodes that were vanced economies declined.5 characterised by high outflows of funds from emerging market economies, not only the vol- … with a In addition to their volumes, the composition ume, but also the rapid speed of the outflows changed of capital flows also changed significantly in from the emerging market economies over the structure of capital flows the aftermath of the global financial crisis (see past few months was especially noteworthy the chart on p. 56). For example, their structure (see the chart on p. 57).9 reveals a decline in cross-border loans, especially from banks.6 This applies mainly to ad- The respective capital flows are determined by vanced economies, but also, to a lesser extent, differing underlying factors, which are trad- to some of the G20 emerging market economies.7 In several emerging market economies, 4 The degree of financial openness – which is calculated as the withdrawal of foreign banks was offset by the sum of a country’s external claims and liabilities divided increasing significance of the bond market. by its gross domestic product – has risen significantly faster among the G20 emerging market economies than the G20 Bank loans and portfolio investment now fund advanced economies since the global financial crisis. emerging market economies in roughly equal 5 Although some individual countries, such as China, account for a very large share of capital flows, this article measure. does not discuss specific countries in detail. 6 See Buch and Goldberg (2020). 7 Based on the Consolidated Banking Statistics of the Bank Sharp decline in The extent to which the COVID-19 pandemic of International Settlements. gross capital has changed global capital flows cannot be 8 See Koepke and Paetzold (2020). flows expected 9 The EPFR data on inflows to equity and bond funds used due to COVID-19 precisely assessed at present due to the lagged here refer to net inflows. Deutsche Bundesbank Monthly Report September 2020 56

Gross capital flows by type of capital flow*

As a percentage of the GDP of the respective group of countries, quarterly data

Gross capital flows to and from G20 advanced economies + 8 Gross capital inflows Other investment Gross capital outflows Portfolio investment in debt + 6 Portfolio investment in equity Net capital flows Direct investment + 4

+ 2

– 2

– 4

– 6 Gross capital flows to and from G20 emerging market economies + 2

– 2

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Source: IMF Balance of Payments. * Gross capital inflows correspond to the balance of purchases less sales of domestic assets by non- residents; gross capital outflows correspond to the balance of purchases less sales of foreign assets by residents. Deutsche Bundesbank

Capital flows itionally categorised into “pull” and “push” fac- While domestic factors, such as growth differ- The volatility of have different tors. Pull factors comprise influencing factors entials, can exert a considerable influence on capital flows is underlying driven largely by factors that originate from the recipient country and the volume of capital flows in the recipient global factors affect investors, while push factors are the country, global factors, such as risk aversion or causes that prevail in the capital-exporting the US monetary policy stance, appear to have country.10 In addition, more recent approaches a larger impact than domestic factors on the examine what are known as “pipes”, which volatility of capital flows in emerging market reflect structural factors in the international economies.13 monetary system.11

Portfolio invest- These factors may exert a stronger or weaker Global financial spillover ment and bank influence depending on the country and situ- loans are more effects likely to be sub- ation. Portfolio investment, like other invest- ject to short- Spillovers term influences ment, appears to be based more on short-term The global financial crisis demonstrated that considerations. It exhibits a significantly nega- financial crises and the risk of contagion are increasingly occur between tive correlation with the level of risk aversion not problems that affect only emerging market emerging market economies and interest rates in the capital-exporting coun- economies. As open economies are signifi- and advanced try. However, it appears to be increasingly influ- cantly interconnected in financial and real eco- economies, too enced by the prevailing domestic fundamentals and the local risk situation in the recipient 10 See Calvo et al. (1996). country. This suggests that investors are differ- 11 See Carney (2019). entiating more clearly between individual 12 See Ahmed et al. (2015). 13 See Bussière et al. (2016), Cerutti et al. (2015) and Pagli- emerging market economies.12 ari and Hanan (2017). Deutsche Bundesbank Monthly Report September 2020 57

nomic terms, shocks – i.e. unexpected events – Inflows to G20 emerging market funds in one country can also have an impact on during various periods of crisis*

other countries, which can in turn have feed- US$ billion, cumulative back effects on the country in which the shock Global financial crisis (t = July 2008) originated; in the literature, these effects are Taper Tantrum (t = May 2013) Chinese stock market turbulence (t = June 2015) known, respectively, as “spillover” and “spill- COVID-19 (t = March 2020) back”. To an increasing degree, this also ap- 0 plies to the relationships between advanced – 10 economies and emerging market economies.

– 20 Spillover effects The spillover of shocks can occur through dir- are driven by a ect and indirect trade channels as well as – 30 variety of inter- connections through direct and indirect financial channels. between the – 40 real economy Direct financial linkages are the result of the and financial cross-border claims and liabilities of banks, – 50 system other financial institutions, governments, en-

terprises and households, which can invest or – 60 obtain funding abroad via loans, direct investment or portfolio investment. Indirect financial – 70 t – 1 t t + 1 t + 2 t + 3 t + 4 t + 5 t + 6 linkages can, for example, arise due to connec- Months tions via third parties, such as common invest- Source: EPFR. * Monthly data on inflows to investment funds ors. The wide range of cross-border intercon- that invest in equity and bond funds in G20 emerging market economies, t = month of crisis, t + 1= month of crisis plus one nectivities and dependencies within the global month. financial system and the real economy are ul- Deutsche Bundesbank timately responsible for the spillover effects between countries and regions. Spillovers can There are two major strands to the literature on occur through a variety of transmission chan- this topic. The first comprises network models nels. In this article, spillover effects are meas- aimed at investigating the causal mechanisms ured without focusing on the channels through of financial contagion. Such analyses are based which the shock is transmitted. The aim of this on balance sheet and macroeconomic funda- analysis is to map the development and mentals. Network models gained in import- changes in the spillover effects over time. This ance following the Asian financial crisis of allows for an assessment of whether, in the 1997-98 and the global financial crisis of light of the growing significance of the emer- 2007-08, with the work of Allen and Gale ging market economies, the mutual depend- (2000) as well as Forbes and Rigobon (2002) ence between the advanced economies and representing seminal contributions.15 the emerging market economies has changed permanently or in specific phases over recent The second strand of the literature is concerned years. with econometric models that use market data to identify spillover effects without making any Greater focus In recent years, the empirical literature has in- further assumptions about the dynamics of on financial creasingly been exploring the issue of cross- shock transmission.16 For example, the ap- spillovers in the literature border spillover effects. While empirical analyses used to look, in particular, at real economic transmission channels such as trade or 14 See IMF (2016) and Bank for International Settlements (2018). commodity markets, increasing financial glob- 15 Comprehensive overviews of existing methods for mod- alisation has lately brought a greater focus on elling contagion effects can be found in Upper (2011), Dun- gey et al. (2005) and Glasserman and Young (2016). analysing financial spillover effects.14 16 See Bricco and Xu (2019). Deutsche Bundesbank Monthly Report September 2020 58

proach of Diebold and Yılmaz, discussed back Germany, with a net spillover index of 32, in the July 2019 issue of the Bundesbank’s sends on average more spillovers to other G20 Monthly Report, centres on a simple and widely countries, while Turkey (-26) and Argentina applicable quantitative measure for spillovers (-17), for example, are on the receiving end of and their evolution over time, revealing trends, spillovers from the other G20 countries. As is to cycles and breaks.17 This method does not be expected, spillovers between countries be- focus on how shocks are transmitted, but longing to the same region, for example in Eur- rather provides a measure for estimating the ope between Germany and Italy (11 from Ger- intensity of spillover effects. The methodology many to Italy and 9 in the other direction) or can thus be applied to measure the strength of between Mexico and Brazil in Latin America spillovers between the G20 countries. (8 from Mexico to Brazil and 7 vice versa) are typically larger than those between two coun- At the same time, the approach offers a way tries from different regions, such as between of assessing spillover effects in the current Germany and Brazil (from Germany to Brazil 6, COVID-19 pandemic (see p. 62). 4 vice versa). At 62.5, the spillover index measuring the average spillovers between the coun- Estimating spill- The outbreak of the pandemic saw share prices tries of the G20 indicates that a large propor- overs between tumble heavily across all G20 countries – in- tion of the variance of all variables is attribut- the equity markets of the G20 cluding countries where case numbers had able to spillover effects from other variables. member states hitherto not been quite as high as elsewhere. In the first half of 2020, the German stock mar- Since the focus is on looking at the countries as Spillovers within ket index DAX and the Euro Stoxx 50 dropped groups, the next step is to analyse the average the emerging market econ- by more than 35% at the peak of the slump spillovers between the emerging market economies and advanced econ- and the US stock market index S&P 500 fell by omies and the advanced economies. A simul- omies groupings more than 30% compared with the end of taneous shock in all G20 emerging market appear more pronounced 2019. Stock market price data are available at economies or all G20 advanced economies is than between high frequency and respond quickly to news, assumed (for more, see the bottom of the box them so the spillover index described earlier is calcu- on p. 62). It is apparent at aggregated level, lated for the yields of G20 benchmark stock in- too, that spillovers within the two country dices (see the box on pp. 59 ff.). These are daily groupings are larger than spillovers from emer- price data in local currency for the period from ging market economies to advanced econ- January 1999 to August 2020. In order to con- omies and vice versa. Moreover, averaged over trol for the different time zones, the yield on the complete analysis period from January each individual stock index is calculated from 1999 to June 2020, advanced economies trans- Friday to Friday. A spillover index value of 100 mitted more spillovers to emerging market indicates that the total variance of a variable is economies than they received from them, with attributable to spillovers of shocks in other vari- the spillover index reaching 42. However, at 29, ables. A value of zero, on the other hand, sig- the spillover index from emerging market econ- nals an absence of discernible spillovers. omies is not negligible.

A country transmits more spillovers than it re- Repeating the analysis for rolling windows of Applying a ceives when the difference between its index 100 weeks each (i.e. roughly two years) makes dynamic analysis … of spillovers to other countries and its index of it possible to represent the spillover matrix for spillovers from other countries is greater than all G20 countries and the group spillover matrix zero. In the literature, this is also referred to as in chart form. The value of the spillover index the net spillover index.18 The analysis suggests that certain countries tend to act more as a 17 See Deutsche Bundesbank (2019). sender rather than as a recipient of spillovers: 18 See Diebold and Yılmaz (2015). Deutsche Bundesbank Monthly Report September 2020 59

The Diebold and Yılmaz spillover index

Unlike other methods based on macroeco- units of time. t denotes the error term not nomic or balance sheet data, which are explained by theε model, where ( ) iid( ) t t ε ≥0~ 0, usually published only a few times per year, and E t, t t t . In the analysis the Diebold and Yılmaz (2009, 2012, 2015) of G20[ε benchmarkε ] = 0 equity ≠ index5 returns, an approach to estimating spillover effects autoregressive lag of p is selected.6 uses daily market data. This enables their = 2 approach to produce a high- frequency The p is then transformed into a measure of spillover effects that adapts “movingVAR( average”:) 7 more quickly than other indicators to 1 1 changes in the data, lending this spillover yt = Ah"t h . − measure some of the greatest predictive h=0 X power amongst existing indicators.2

Since the Diebold and Yılmaz spillover index also requires only minimal assumptions, this 1 See Diebold and Yılmaz (2009). Diebold and Yılmaz methodology is employed in a wide variety argue that, with increasing data frequency, an empirical model is better able to correctly track the time of papers. It can be applied to price- based variation of spillover effects. In addition, for certain analyses as well as to quantitative variables countries – including some of the G20 emerging market economies – it is diffi cult to obtain reliable data on and can also be used to model potential the macroeconomic fundamentals and the balance transmission channels. In this box, the ap- sheets of general government, fi nancial institutions and enterprises. proach will be used to estimate fi nancial 2 See Arsov et al. (2013). Moreover, the spillover meas- spillovers based on equity market returns of ure is closely related to other known systemic measures of risk such as the CoVaR, introduced by Adrian the benchmark indices of G20 countries. and Brunnermeier (2016), and the Marginal Expected Shortfall, published in Acharya et al. (2016). 3 Here and in the rest of this box, the focus is on the The spillover index is based on a rolling esti- countries belonging to the Group of 20 (G20). The mate of VAR models in which the variances European Union, which is also a G20 member, is disregarded. In this box, the G20 emerging market econ- of the forecast errors are decomposed. On omies comprise Argentina, Brazil, China, India, Indo- this basis, a time-varying index is then con- nesia, Mexico, Russia, Saudi Arabia, South Africa and Turkey, and the advanced economies in the G20 are structed. Australia, Canada, France, Germany, Italy, Japan, South Korea, the United Kingdom and the United States. 4 The VIX is additionally used in the estimate in order The estimate for a single time window then to control for global risk aversion, which could be a follows the reduced- form p model: powerful driver of data patterns. The VIX serves here VAR( ) as a sort of “quasi-exogenous” variable that is omitted p when calculating the spillover indices. The analysis is also rerun as a VARX model with the VIX as an exogen- yt = hyt h + "t , ous variable, which leads to similar results. h=1 X 5 The benchmark equity indices used here are MERVAL, S&P/ ASX 200, Bovespa, S&P/ TSX, Shanghai SSE, DAX, CAC 40, FTSE 100, IDX Composite, Nifty50, FSTE MIB, where yt is a vector with observations of all Nikkei 225, KOSPI, S&P/ BMV IPC, MOEX, Tadawul All N endogenous variables. In the present Share, S&P 500 and JSE, and they are called up via Bloomberg. The vast majority of these are price in- case, these are the daily returns of the bench- dices. Only the DAX and Bovespa are performance in- mark indices3 of the G20 countries.4 rep- dices. The selected benchmark indices are used in this Φ fashion in other research papers, too. resents an N ×N matrix with regression co- 6 Based on the Akaike Information Criterion (AIC), effi cients that refer to the observations of p is selected. 7 =See 2 Lütkepohl (2005) and Kilian and Lütkepohl the endogenous variable (yt-h) lagged by p (2017). Deutsche Bundesbank Monthly Report September 2020 60

Here, the N ×N coeffi cient matrix Ah fol- GIRFs model the dependent variables’ re-

lows the recursion Ah Ah Ah sponses to shocks to each variable in the = Φ1 –1 + Φ2 –2 pAh p where A represents an system. A shock with a magnitude of one + … + Φ - 0 N ×N identity matrix. In addition, Ah standard deviation to each error term is (if h ) . Accordingly, in the “moving aver- = 0 simulated separately on each equation, pro- < 0 age” representation, the current value of a ducing a total of N 2 GIRFs. variable is defi ned via a function of its current and past error terms. In the next step, the impulse response functions are used to calculate the fore- In the next step, impulse response functions cast error variance FEV for each vari- are created in order to estimate the time able. It gives the dispersion( ) of each vari- profi le of a shock which hits the system at able which, owing to the shock that δ time t up to time t H given the absence occurred in t , would have been impossible of any other shocks +to the system. The fore- to forecastε between t and t H. For 8 + casting horizon H is chosen here. In- the variable yi the FEV is given as = 10 H 1 stead of the Cholesky decomposition, the FEV (y ! )= − e0 A A0 e . A i,t+H | t h=0 j h h j generalised VAR framework of Koop, Pesa- shock to the variableP j directly impactsP the ran and Potter (1996) and Pesaran and Shin variable itself but, due to the dynamic struc- (1998) is used. The advantage of the gener- ture of the VAR model, can also affect all alised framework is that no assumptions re- other variables in the system. The contribu- garding the causal relationships between tions by each of the individual shocks to the the disturbance terms are necessary. How- FEVs of the respective variables can be cal- ever, the downside is that contemporan- culated using a forecast error variance de- eous causal effects cannot be modelled.9 composition FEVD . The FEV of variable ( ) Instead of shocking all elements of t, only i (H steps ahead) explained by a shock in ε the j th element in t is shocked, and the the equation of variable j is calculated as: impacts of other shocksε are disregarded as- 1 H 1 2 − − (e A e ) suming multivariate normal distribution of jj h=0 j0 h j ⇥ij (H)= H 1 . h=0− (ei0Ah Ah0 ei) t. The shock j in the variable j,t of one P P 1 ε δ 2 ε standard deviation jj at time t creates a P P generalised impulse response function The results for all i variables and j shocks (GIRF) of: can be represented in an N ×N matrix

where the element ij represents the share GIRF (h)=E[yt+h "j,t = δj, !t 1] Θ | in the FEV of variable i explained by a E[yt+h !t 1] 1 | 2 = σjjAh⌃ej

8 The ten-step- ahead forecasting horizon is a standard where h H is the forecast hori- assumption in the literature. See Diebold and Yılmaz 1 zon, 2 the= 0,standard 1, …, deviation of the error (2009). jj 9 Alternative approaches generally require additional term of the j th equation, Ah the coeffi cient assumptions. Although a Cholesky decomposition matrix, the covariance matrix of the leads to orthogonal shocks and thus allows for un- equivocal identifi cation, it also requires additional as- error term t and ej a selection vector of di- sumptions concerning the contemporaneous causal mension εN where the j th element relationships between the variables. The approach pro- ( ×1) posed by Bettendorf and Heinlein (2019) presented in takes a value of one and all other elements the July 2019 issue of the Bundesbank’s Monthly Re- take a value of zero. comprises all in- port (see Deutsche Bundesbank (2019)) likewise pre- t sumes the existence of a clear causal structure be- ω( –1) formation known up until time t . The tween the error terms. ( –1) Deutsche Bundesbank Monthly Report September 2020 61

Spillover matrix

Index of spillovers to country i from countries i j Country 1 Country 2 … … Country N j i → ≠ Country 1 ~ ~ … … ~ N ~ j j Θ11 Θ12 Θ1 Θ1← , ≠1 Country 2 ~ ~ … … ~ N ~ j j Θ21 Θ22 … … Θ2 Θ2← , ≠2 … … … … … … … … … … … …

Country N ~ N ~ N … … ~ NN ~ N j j N Θ 1 Θ 2 Θ Θ ← , ≠ Index of spillovers from country j to countries i j ~ i i ~ i i … ~ i N i N SI ≠ Θ ←1, ≠1 Θ ←2, ≠2 Θ ← , ≠ Deutsche Bundesbank

10 shock to variable j. Element AR,DE, for in- This measure can also be used to measure stance, would be the share inΘ the FEV of directional spillovers from variable j to all the return of Argentina’s equity price index other variables i N: attributable to shocks in the model equa- = 1, …, N tion for the return of the German DAX. ⇥i j i=1,i =j ⇥ j = 6 , Since, in the generalised framework, shocks ⌅ N P e in each variable are not necessarily orthog- e onal, the shares in the FEV of a variable and, by analogy, directional spillovers to explained by shocks do not necessarily add variable i from all variables j N : up to 100%. Accordingly, each element of = 1, …, N the FEVD matrix is normalised with the re- j=1,i =j⇥i j ⇥ij ⇥ = 6 ⇥ = i ⌅ . spective row total, i.e. ij N ⇥ , N j=1 ij P N e which means that ⇥ P and j=1 eij e N = 1 ⇥ij N. The FEV of each variable For each variable in the system, the direc- i,j=1 P e in the system= is thus equally weighted, tional “from” and “to” spillover indices can P e which means that the rows of the normal- be used to calculate a net spillover index ised matrix can be compared with one an- that shows whether a variable is more likely other. This enables the variance shares in to be the source or recipient of spillover ef- both the rows and the columns of the nor- fects. The normalised FEVD matrix, the malised matrix to be added up and com- directional spillover indices, the net spillover pared for different variables. index and the total spillover index are presented in a spillover matrix (see the table Off- diagonal entries in the normalised above).

FEVD matrix ~ ij where i j are used as a measure of spillover(Θ effects ≠ between) the To gain an overview of the development variables in the system. In order to obtain a and intensity of spillovers between ad- spillover index SI for the estimated period, the total of all off-diagonal entries is divided 10 In the generalised framework, causal spillovers – in by the total of all entries in the matrix: the sense that spillover effects are only propagated from one given country to another country – can be N N modelled only using lagged variables. The results i,j=1,i =j⇥ij i,j=1,i =j⇥ij would be skewed accordingly if such effects material- 6 6 SI = N = . ised and spilled over at the same time (see Bettendorf ⇥ij N P i,j=1 e P e and Heinlein (2019)). P e Deutsche Bundesbank Monthly Report September 2020 62

vanced economies and emerging market and, conversely, the index of spillovers from economies, a measure of a group spillover advanced economies to emerging market index is formed. This is done by assuming a economies as: 1 systemic shock in all countries within a EME i EME i AE ⇥ij ⇥EME AE = 2 2 . group G (in this case, the nine advanced G P P e economies of the G20, AE, and the ten e G20 emerging market economies, EME ) The spillover matrix described above and and then analysing their combined impact the reduced-form group spillover matrix are on all countries outside the group. When estimated repeatedly for rolling time win- looking at the group as a whole, a spillover dows of length w in order to identify the matrix in which the row total for each dynamics of the spillover effects over time. of the individual entries equals one is The T w spillover matrices which this pro- additionally created by transforming the duces can– be displayed in spillover plots. G i AE j=1 ⇥ij 2 entries into ⇥AE j = AE and P P G e i EME j=1 ⇥ij ⇥EME j = 2 . e EME P P e Thee index of spillovers from emerging market economies to advanced economies is given as: 1 AE i AE i EME ⇥ij ⇥AE EME = 2 2 G P P e e

for a given point in time is calculated on the (black line) is lower than that between all G20 … reveals an basis of the corresponding window covering countries (grey dashed line). This is because the increase in spillovers during the last 100 weeks.19 latter also reflects spillovers between advanced periods of stress economies amongst themselves and emerging The group spillover index between emerging market economies amongst themselves, and market economies and advanced economies these tend to be relatively high. For the pur- poses of the grouped perspective, these are, Index of spillovers between emerging however, included in the respective own share market economies and advanced for the grouping – i.e. the diagonal entries in economies in the G20* the group spillover matrix. G20 benchmark equity indices, average values, January 1999 to August 2020 The adjacent table shows the indices of group Very significant from … spillover (the non-diagonal entries) from emer- rise in spillover emerging effects during market advanced Spillover ging market economies to advanced econ- the COVID-19 Item economies economies index pandemic Emerging market omies and vice versa. These are plotted in the economies 58 42 . chart on p. 63 as light blue and dark blue lines. Advanced economies 29 71 . This dynamic representation allows an assess- Spillover index . . 36 ment of how spillovers have evolved over time. Sources: Bloomberg and Bundesbank calculations. * For the pur- poses of this analysis, the G20 emerging market economies One can see that spillover effects have in- comprise Argentina, Brazil, China, India, Indonesia, Mexico, Russia, Saudi Arabia, South Africa and Turkey, and the advanced economies in the G20 are Australia, Canada, France, Germany, 19 Note that the calculations discounted points in time for Italy, Japan, South Korea, the United Kingdom and the United States. which there were no data for one or more variables. This means that any 100-week window refers to the last 100 Deutsche Bundesbank weeks when data were available for all variables. Deutsche Bundesbank Monthly Report September 2020 63

Spillover index* for G20 benchmark equity indices

%, weekly data, averages calculated on a rolling basis over the preceding 100 weeks

100 100

Between all G20 countries1 90 90

80 80

70 70

60 From emerging market economies to advanced economies 60 From advanced economies to emerging market economies Between advanced economies and emerging market economies 50 50

40 40

30 30

20 20 2000 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 J FMAM J J 2020

Sources: Bloomberg and Bundesbank calculations. * Spillover indices measure the strength of spillover effects. They are calculated by performing a generalised forecast error variance decomposition in a VAR(2) model with a ten-day forecast horizon and row normalisa- tion based on Koop et al. (1996), Pesaran and Shin (1998) and Diebold and Yılmaz (2009, 2012). The spillover indices describe the percentage of the variance explained by foreign shocks in the respective periods. 1 Includes spillover effects from advanced economies to emerging market economies and vice versa but also spillover effects within the group of advanced economies and that of emerging market economies – for instance, spillovers from Germany to France. Deutsche Bundesbank creased especially when financial markets have have fundamentally changed the conditions Crises in individual experienced phases of heightened stress, such for national economic policy in open econ- emergıing market economies can as the bursting of the dotcom bubble in 2000, omies. spread to the global financial crisis and the onset of the others … COVID-19 pandemic. These final two events In particular, as shown above, countries are vul- … and can also represent the strongest and, above all, most nerable to shocks or spillovers emanating from exert spillback effects on rapid rise in spillovers in the period covered by other countries. In the past, crises in one coun- advanced the analysis. That the COVID-19 pandemic is so try have already been known to spread to other economies clearly reflected in the estimation results is strik- emerging market economies or induce wake- ing. Having said that, compared with other cri- up calls for investors.20 With emerging market sis periods, this pandemic has hit all countries economies now more deeply integrated into hard and, most importantly, unexpectedly. It the global financial system, this could also in- remains to be seen how spillover effects will creasingly entail spillbacks to advanced econ- evolve over time and whether a prolonged omies. period of high spillovers will set in, as it did in the wake of the global financial crisis. It is also conceivable that, where investors harbour mistrust towards a country with high levels of debt and pursuing economic policies Vulnerabilities in emerging that do not inspire confidence, such sentiments market economies may spread to other countries with similar issues, be they an emerging market economy The rise in international capital flows and the rapidity with which capital markets respond 20 See Ahnert and Bertsch (2015). Deutsche Bundesbank Monthly Report September 2020 64

A heatmap for the external stability of selected emerging market economies

Similarly to international institutions, the rate against the US dollar ( S),4 the per- Bundesbank calculates and assesses a var- centage change in foreign ∆%reserves ( R) iety of metrics that shed light on the exter- and the difference in short- term domestic∆% nal stability of selected emerging market interest rates relative to the short- term economies (EMEs).1 These include a metric interest rate level of the United States for assessing a country’s foreign currency ( I):5 reserve adequacy (ARA), exchange market ∆ pressure (EMP), and infl ows from invest- 1 EMP i,t = (Ii,t %Si,t %Ri,t). ment funds that invest in EMEs. Warning 3 signals can be derived from these indicators in the event that certain thresholds are ex- A warning signal is triggered when the ceeded or undershot. By colour- coding the value of the EMP indicator exceeds its long- number of such warning signals for each term, country-specifi c average by 1.5 times country and each point in time, they can be the standard deviation of the index.6 presented visually in the form of a heatmap. Overall, the EMP indicator is less persistent The three selected indicators and the defi n- than the ARA indicator. Warning signals ition of the warning signals are described in thus only appear for relatively short periods more detail below. of time in the EMP indicator. One reason for this is that in contrast to the ARA indicator, The ARA indicator shows whether a country the EMP indicator is not based on levels, has suffi cient reserves to compensate for an but rather on rates of change (within a outfl ow of foreign currency in the event of quarter). The latter generally revert quickly a temporary crisis. The magnitude of poten- to their average values. The variables exam- tial outfl ows is determined on the basis of historical experience (i.e. from past balance 1 The EMEs are a selection of those represented in the of payments crises) and current macroeco- G20 (Argentina, Brazil, China, India, Indonesia, Mex- nomic conditions. If the existing foreign ico, Russia, South Africa and Turkey). In addition to this, the Bundesbank also studies individual countries currency reserves are lower than the thresh- that are currently subject to especial external exposure old value derived from them, a warning sig- on account of particular events or that are in the spot- light for any other reason on an ad hoc basis. nal is emitted in the corresponding quarter. 2 See IMF (2011, 2015) and Deutsche Bundesbank The indicator was originally developed by (2017). 3 See Hossfeld and Pramor (2018). the International Monetary Fund (IMF). The 4 Indirect quotation: an increase in the exchange rate Bundesbank uses a slightly modifi ed ver- represents an appreciation of the respective domestic currency. sion.2 5 The indicator based on the unweighted average has proven to be a particularly robust measure in empirical studies. The EMP indicator measures the exchange 6 The EMP indices are adjusted for statistical outliers market pressure affecting a country’s cur- when calculating the standard deviation. Without such adjustment, particularly severe crises would distort the rency (i) at time t. Various options for esti- standard deviation upwards to such an extent that mating the EMP indicator are discussed in smaller crises would no longer be identifi ed. An outlier is defi ned as an absolute value that exceeds the the literature.3 The indicator used here is country- specifi c absolute minimum of the EMP index derived from the unweighted average of by 100. This threshold value was chosen so that known crisis periods from the past could be identifi ed the bilateral rate of change in the exchange as successfully as possible. Deutsche Bundesbank Monthly Report September 2020 65

ined by the EMP indicator would thus have Heatmap for warning signals regarding to experience above- average change over a the external stability of emerging market relatively long period of time to trigger a economies' capital flows signal lasting over several quarters. By contrast, if holdings of foreign currency re- Number of warning signals 0 1 2 3 serves fall under the threshold value and Argentina remain there, this is enough to trigger a Brazil persistent signal in the ARA indicator. China Indonesia India The infl ows from investment funds that in- Mexico vest in a specifi c country serve as a timely Russia Turkey indicator for the gross capital fl ows into South Africa 7 that country. A warning signal is triggered 2017 2018 2019 20 as soon as the sum of investment in debt Deutsche Bundesbank securities and equity instruments demon- strates that international investors have withdrawn funds from a country over one timately mean that the critical mass of for- quarter. In this context, it is rather unusual eign reserves is undershot. for the countries under observation to dis- play an outfl ow of funds exceeding a period The three indicators can be summarised in a of three months. The threshold value is heatmap, which paints a clear yet nuanced therefore selected to ensure that even mar- picture of a country’s external vulnerability. ginal outfl ows trigger a warning signal. This The total number of warning signals from indicator correlates strongly across the indi- the various indicators is added up for each vidual EMEs, suggesting that investors tend point in time and each country. The corres- to make their investment decisions for sev- ponding sum is then colour- coded to pro- eral EMEs at the same time. That said, vide a general overview. For instance, a country- specifi c factors, which may even country’s cell is shaded light grey if no occasionally outweigh cross- border infl u- warning signal appears at the correspond- ences, also play a role. ing point in time across all indicators. If there is a single warning signal, the cell is Each of the three indicators can be studied shaded light blue. If there are two or three individually in order to identify tensions in a warning signals, the cell is shaded medium specifi c area of international capital fl ows. blue or dark blue respectively. A dark blue However, severe tensions often appear in signal is thus interpreted to mean that, at several different indicators at the same the point in question, a country’s foreign time, as the individual indicators are not in- currency reserves are too low, it is subject dependent of each other. If, for example, to exchange rate pressure and, at the same international investors shift their focus away from a particular investment location, the effect of this is unlikely to be restricted to 7 The data are based on the monthly reports of EPFR Global. These reports depict the infl ows from invest- capital outfl ows. The currency of the af- ment funds that submit information to EPFR Global. fected country may also come under pres- The values can therefore only be taken as an approximation of the offi cial balance of payments data. As the sure, resulting in a persistent outfl ow of for- offi cial data are generally only published after a signifi - eign reserves if the central bank attempts to cant delay, developments in international capital fl ows can be assessed considerably sooner with the data bolster the national currency. This may ul- used here. Deutsche Bundesbank Monthly Report September 2020 66

time, international investors are withdraw- refl ects the fact that the warning signals ing funds from that country. and indicators are, in part, based on strong assumptions and are subject to estimation The heatmap (see the chart on p. 65) shows uncertainty. For example, a stable relation- the results for the EMEs of the G20. The ship between the observable variables dur- period under review runs from the fi rst ing past and possible current crises is as- quarter of 2017 to the fi rst quarter of 2020. sumed. Even summarising the indicators by giving all warning signals the same weight- The chart clearly illustrates the increase in ing represents a considerable simplifi cation. external vulnerability associated with the Ultimately, the three indicators under con- COVID- 19 pandemic in the fi rst quarter of sideration deliver only a limited overview of 2020. On account of the pandemic-related the developments in international capital uncertainty regarding future global eco- movements. Nevertheless, the results pro- nomic developments, investors appear to vide key indications as to which countries’ have withdrawn capital from EMEs to a capital fl ows merit closer analysis with re- greater extent. This resulted in the relevant gard to external stability. indicator showing a warning signal for all countries under observation during the fi rst three months of 2020, and also explains why there is at least a light blue signal on the heatmap for each country.

As the estimated ARA indicator in the fi rst quarter of 2020 also confi rms an insuffi - cient stock of foreign currency reserves in the case of Argentina, Brazil, Mexico, Tur- key and South Africa, these countries dis- play at least a medium blue signal.

The heatmap even shows dark blue signals for two of the countries under observation. This can be attributed to the fact that heightened exchange market pressure could be observed for Argentina and Brazil in addition to the aforementioned warning signals.

A heatmap can be used to depict key developments and relationships in a timely and concise manner. Even so, it should always be embedded into a more comprehensive analysis rather than serving as a replace- ment for this; in spite of the extensive analyses behind the individual indicators, external sector developments are ultimately presented in a highly simplifi ed manner. This Deutsche Bundesbank Monthly Report September 2020 67

or an advanced economy. The magnitude of By implementing appropriate policy measures, the distortions experienced in the countries is emerging market economies can help ensure therefore contingent not only on the external that capital inflows do not have a destabilising impulse but also on the domestic vulnerabilities effect and instead bring sustainable economic of the countries. benefits. Well-developed and regulated financial systems and, above all, sound macroeco- Vulnerabilities in The box on pp. 64 ff. takes a closer look at the nomic policies can bolster resilience; they can G20 emerging external vulnerability of emerging market econ- help to prevent crises and to give economic market economies signifi- omies belonging to the G20. Analysing past policymakers room for manoeuvre when chal- cantly higher in some cases as financial market and balance of payments cri- lenging economic conditions arise. a result of ses reveals that exchange market pressure on COVID-19 the domestic currency, foreign currency reserve A potential starting point in terms of expanding Developing local holdings and the behaviour of international in- this scope for economic policy action lies in de- capital markets can lessen vestors are key factors. Critical values are deter- veloping local capital markets. This would im- vulnerability to mined for these three indicators; whenever they prove access to borrowing in local currency, crises exceed or fall below these markers, a warning and therefore reduce financing needs in for- signal is triggered. The investigation shows that eign currency for a given debt level, which is the external stability of certain countries deteri- considered one of the major reasons behind orated sharply during the COVID-19 pandemic. heightened financial system vulnerability. This This is particularly true of Argentina and Brazil, has prompted many emerging market econ- which breached the predefined thresholds for omies to push forward with the development all three criteria. This does not necessarily mean of local capital markets in recent years.21 Al- that the other countries are less at risk of exter- most everywhere, this has been accompanied nal imbalance, however. Examining the selected by a significant increase in the proportion of indicators and the warning signals triggered local currency debt, which shifts the exchange can only furnish initial reference points for a rate risk for that portion of debt to foreign more in-depth analysis. This is partly because it creditors. While issuance in local currency has is not possible to take all potential threats into become established practice for government account. Furthermore, the threshold values in securities, corporate debt continues to be pre- question were calculated on the basis of histor- dominantly denominated in foreign currency. ical crisis patterns and with transmission mech- Moreover, events in spring 2020 showed that anisms assumed to hold constant. These may markets for debt instruments denominated in change over time, however. domestic currency are also exposed to heightened volatility not least due to international investors selling off debt instruments issued by Economic policy options for emerging market economies.22 Since foreign utilising the benefits of investors’ investment motives need not be cross-border capital flows geared only to local circumstances but also to regional developments or to the behaviour of Appropriate Capital flows to emerging market economies other investors, building up a sufficiently large economic policy have, in recent decades, been highly volatile: base of domestic investors might be helpful in can influence the impact of periods of strong inflows have been followed capital flows by times where inflows have suddenly dried up 21 This has also been accompanied by a significant in- or where there have even been massive out- crease in debt, however. According to the BIS, public debt flows. The forces driving these phenomena on the capital market rose across a broad base of emerging market economies from 18% of gross domestic product in may lie in the emerging market economies 2000 to 37% in 2017. See Wooldridge (2020). themselves, but may also be the product of 22 This appears to be due to the presence of a small number of large investment funds in these countries and a ten- global trends. dency toward herd behaviour. See Wooldridge (2020). Deutsche Bundesbank Monthly Report September 2020 68

limiting volatility from this source. However, In addition to the instruments mentioned Macroprudential emerging market economies often still lack the above, countries may also implement measures measures may also influence supporting infrastructure, such as liquid deriva- to influence capital flows more directly. This capital flows and are often tives markets, which facilitate local trading. category covers macroprudential policies but implemented also capital controls. Since the financial crisis, When capital flows exhibit a high degree of macroprudential measures have increasingly volatility, flexible exchange rates can, in prin- been employed as an economic policy tool and ciple, absorb some of the external shocks. are mostly used to influence credit growth and Monetary policy can also respond, with the financial institutions’ leverage by curbing or fa- classic instruments being interest rate adjust- cilitating the inflow of capital. They are often ment or interventions in the foreign exchange used as preventative measures to stop imbal- market. Recent IMF analyses suggest that, con- ances from emerging in the financial system, trary to previous assumptions, the transmission and typically tend to be long-term. Despite of interest rate policy decisions in emerging their growing use, however, there has so far market economies which have adopted infla- been little empirical evidence on the extent to tion targeting can be as effective as in ad- which macroprudential measures are able to vanced economies.23 reduce spillover effects.25 The easing of macroprudential measures is one of the most fre- Foreign Sufficiently large foreign exchange reserves can quently used economic policy responses to the exchange increase the options available to emerging mar- COVID-19 crisis alongside fiscal and monetary reserves are an important ket economies. First, foreign exchange reserves policy responses. Almost every country made tool in avoiding crises built up in good times have an important posi- use of them in one form or another with the tive signalling function, meaning that, in a primary objective of facilitating access to liquid- best-case scenario, the mere existence of high ity across the individual G20 member states.26 reserves can prevent investors from losing confidence and a crisis from breaking out. Second, Macroprudential measures may also take the Capital controls solid reserve buffers can be used to temporarily form of capital controls when they affect cap- are often subject to criticism, … offset capital outflows in the event of a crisis ital flows. Academic literature and international and thus limit excessive exchange rate fluctu- organisations are largely critical of such meas- ations. However, interventions of this type ures due to their distortionary effects on capital should only function as an adjustment mech- allocation. However, those taken as part of a anism during periods of heightened volatility in longer-term strategy to liberalise capital move- capital flows. They should not be used to re- ments are largely undisputed. They should be place necessary economic policy corrections or used to prevent imbalances from building up as prevent fundamentally necessary adjustments long as the financial systems in question remain to the (real) exchange rate in the long term. relatively underdeveloped.

Isolated foreign Given the current economic turmoil, it is diffi- As experience of the global financial crisis has … but they may exchange cult to unequivocally assess the intervention shown, capital controls are increasingly being have a positive market interven- impact on finan- tions during the behaviour of most countries as these data are seen as a potential means of safeguarding cial stability COVID-19 pandemic often confidential. According to the IMF, indi- financial stability. For example, they could be vidual countries such as Brazil, Russia, Turkey used to change the composition of capital and Indonesia have intervened in the foreign exchange market on a number of occasions 23 See Brandao-Marques et al. (2020). since February 2020 or, like India, have carried 24 See IMF (2020a). In the case of Russia, sales of foreign currency from the National Welfare Fund on 10 March out foreign exchange swaps.24 2020 are due to the oil price falling below the reference value. See also IMF (2020b). 25 See Buch and Goldberg (2020). 26 See IMF (2020a). Deutsche Bundesbank Monthly Report September 2020 69

flows in favour of capital flows with low volatil- During the COVID-19 pandemic, the IMF has ity. In the case of critical capital outflows, cap- already provided financial support to over 85 ital controls could also be regarded as a legit- member countries and has taken a number of imate policy instrument under certain condi- measures to assist members more effectively. tions. This thinking is outlined in the Institu- For example, the access limits for emergency tional View,27 authored by the IMF in 2012, credit (Rapid Credit Facility (RCF) and Rapid which considers the deployment of capital con- Financing Instrument (RFI)) and the annual trols as a possible policy option depending on access limits for financial assistance were tem- country-specific considerations. Owing to the porarily increased. In addition, the priorities of side effects associated with this instrument, for the IMF work programme and internal pro- instance with regard to circumventions, meas- cesses have been revised in order to enable the ures regulating capital movements should be Fund to respond more quickly to members’ re- transparent and temporary. As soon as the crit- quests for assistance. In addition, a new tem- ical situation is over, these controls should be porary short-term liquidity line was set up for lifted. Under no circumstances should they be members with very strong economic funda- used to delay necessary macroeconomic ad- mentals in order to cushion moderate balance justments. According to the IMF, no G20 coun- of payments needs arising due to tension in the try has yet introduced additional capital con- international capital markets. trols in the context of the current COVID-19 pandemic. Conclusion International The grave global consequences of the COVID- community can 19 pandemic have once again made it clear The global interconnectedness of national Still not possible provide support that the associated challenges at the national financial systems has continued to increase to assess the COVID-19 pan- level may be beyond the financial capacities of over the past two decades, although this trend demic’s impact on financial less developed countries in particular. These has lessened somewhat since the global finan- market require supplementary international policy cial crisis. It is not possible at this stage to con- integration approaches, policy advice and financial sup- clusively assess how the COVID-19 pandemic port. The latter can be provided through official will affect this trend. Given the high level of financial assistance from countries and organ- global economic integration achieved, emer- isations, the IMF, multilateral and regional de- ging market economies and, increasingly, ad- velopment banks and/or regional institutions vanced economies are vulnerable to external (Regional Financing Arrangements). In add- shocks. Global capital flows are a key transmis- ition, the central banks of reserve currency sion channel, and their volatility can present a countries may, within their mandate, grant challenge, especially to countries with less de- swap lines or repo facilities to other central veloped financial systems. The current COVID- banks in order to safeguard the liquidity of the 19 pandemic has shown once again that, dur- money market in foreign currency.28

27 The full title is The Liberalization and Management of IMF responded Owing to its global membership, its mandate Capital Flows: An Institutional View. See IMF (2012). swiftly and com- and its expertise, the IMF plays a prominent 28 Since the global financial crisis, there has been an ar- prehensively to rangement between the Federal Reserve, the ECB, the Bank the COVID-19 role in helping to combat balance of payments of Japan, the Bank of England, the Bank of Canada and the pandemic problems. With permanent resources of around Swiss National Bank. Following the outbreak of the COVID- 19 pandemic, the ECB additionally agreed on temporary €570 billion, funds from credit lines worth and limited swap lines with Bulgaria, Croatia and Denmark. around €600 billion for crisis situations and The US Federal Reserve set up similar swap lines with Aus- tralia, Brazil, Denmark, Mexico, New Zealand, Norway, additional trust funds for financial assistance to Singapore, South Korea and Sweden. Both the Federal Re- low-income countries, it has ample financial serve and the ECB have also established securities repur- chase agreements with other central banks. See Federal resources to support members when needed. Reserve Board (2020) and European Central Bank (2020). Deutsche Bundesbank Monthly Report September 2020 70

ing periods of stress, a flight to safe invest- exceeded the level of the global financial crisis ments, domestic assets and cash can exacer- at an early stage. bate tensions for emerging market economies. However, the analysis also shows that, as finan- The extent of the vulnerabilities is not deter- cial spillovers to and spillbacks from emerging mined solely by the spillover effects, but also market economies increase, domestic eco- needs to be assessed together with local condi- nomic policy in advanced economies may also tions in the individual countries. This shows be exposed to potentially destabilising influ- that the potential vulnerabilities of individual ences. G20 members have increased markedly in the current crisis. The extent to which these vulner- The G20 countries present a mixed picture in abilities will materialise is partly determined by this respect: some countries tend to transmit each country’s economic policy. The develop- spillover effects on balance, whilst others ment of local capital markets and a sufficient mainly receive them. The effects within country stock of foreign exchange reserves are useful groups are more pronounced than between components of a stability-oriented macroeco- advanced and emerging market economies, nomic policy. From a financial stability perspec- although the advanced economies are still tive, it may make sense to deploy macropru- responsible for stronger spillovers to the emer- dential measures as well as – under certain cir- ging market economies than vice versa. Worth cumstances – measures designed to manage noting is the stark rise in transmission effects at capital flows. Where necessary, the inter- the onset of the COVID-19 pandemic, which national community can also provide support.

List of references

Acharya, V. V., L. H. Pedersen, T. Philippon and M. Richardson (2016), Measuring Systemic Risk, The Review of Financial Studies, Vol. 30, No 1.

Adrian, T. and M. K. Brunnermeier (2016), CoVaR, American Economic Review, Vol. 106, No 7, pp. 1705-1741.

Ahmed, S., B. Coulibaly and A. Zlate (2015), International Financial Spillovers to Emerging Market Economies: How Important Are Economic Fundamentals?, International Finance Discussion Papers 1135.

Ahnert, T. and C. Bertsch (2015), A Wake-Up-Call Theory of Contagion, Staff Working Papers 15-14, Bank of Canada.

Allen, F. and D. Gale (2000), Financial Contagion, Journal of Political Economy, Vol. 108, No 1, pp. 1-33.

Arsov, I., E. Canetti, L. Kodres and S. Mitra (2013), Near-Coincident Indicators of Systemic Stress, IMF Working Paper, 13/115.

Bank for International Settlements (2018), Financial spillovers, spillbacks, and the scope for international macroprudential policy coordination, BIS Papers No 97. Deutsche Bundesbank Monthly Report September 2020 71

Bettendorf, T. and R. Heinlein (2019), Connectedness between G10 currencies: Searching for the causal structure, Deutsche Bundesbank Discussion Paper No 06/2019.

Borio, C. and P. Disyatat (2015), Capital flows and the current account: Taking financing (more) seriously, BIS Working Paper No 525.

Bricco, J. and T. Xu (2019), Interconnectedness and Contagion Analysis: A Practical Framework, IMF Working Paper, October 2019.

Brandao-Marques, L., G. Gelos, T. Harjes, R. Sahay and Y. Xue (2020), Monetary Policy Transmis- sion in Emerging Markets and Developing Economies, IMF WP/20/35.

Buch, C. and L. Goldberg (2020), Global Banking: Towards an Assessment of Benefits and Costs, Annual Review of Financial Economics, forthcoming.

Bussière, M., J. Schmidt and N. Valla (2016), International Financial Flows in the New Normal: Key Patterns (and Why We Should Care), CEPII Policy Brief 2016-10, CEPII Research Center.

Calvo, G., L. Leiderman and C. Reinhart (1996), Inflows of Capital to Developing Countries in the 1990s, Journal of Economic Perspectives, Vol. 10 No 2, Spring 1996, pp. 123-139.

Carney, M. (2019), Pull, push, pipes: sustainable capital flows for a new world order,https://www. bankofengland.co.uk/speech/2019/mark-carney-keynote-remarks-at-the-2019-institute-of- international-finance-spring-membership-meeting

Cerutti, E., S. Claessens and D. Puy (2015), Push Factors and Capital Flows to Emerging Markets: Why Knowing Your Lender Matters More Than Fundamentals, IMF WP/15/ 127.

Committee on the Global Financial System (2009), Capital flows and emerging market economies, CGFS Papers No 33.

Deutsche Bundesbank (2019), Parallels in the exchange rate movements of major currencies, Monthly Report, July 2019, pp. 17-37.

Deutsche Bundesbank (2017), Do sufficient foreign exchange reserve holdings reduce the prob- ability of balance of payments crises?, Monthly Report, October 2017, pp. 23-25.

Diebold, F. X. and K. Yılmaz (2015), Financial and Macroeconomic Connectedness – A Network Approach to Measurement and Monitoring, Oxford University Press.

Diebold, F. X. and K. Yılmaz (2014), On the network topology of variance decompositions: Meas- uring the connectedness of financial firms, Journal of Econometrics, Vol. 182, No 1, pp. 119-134.

Diebold, F. X. and K. Yılmaz (2012), Better to give than to receive: Predictive directional measurement of volatility spillovers, International Journal of Forecasting, Vol. 28, pp. 57-66.

Diebold, F. X. and K. Yılmaz (2009), Measuring financial asset return and volatility spillovers, with application to global equity markets, The Economic Journal, Vol. 119, No 534, pp. 158-171. Deutsche Bundesbank Monthly Report September 2020 72

Dungey, D., R. Fry, B. Gonzáles-Hermosillo and V. Martin (2005), Empirical modelling of contagion: a review of methodologies, Quantitative Finance, Vol. 5, pp. 9-24.

European Central Bank (2020), Press Release: New Eurosystem repo facility to provide euro liquidity to non-euro area central banks, 25 June 2020, https://www.ecb.europa.eu/press/pr/date/2020/ html/ecb.pr200625~60373986e5.en.html

Federal Reserve Board (2020), Press Release: Federal Reserve announces the establishment of temporary U.S. dollar liquidity arrangements with other central banks, 19 March 2020, https:// www.federalreserve.gov/newsevents/pressreleases/monetary20200319b.htm

Forbes, K. J. and R. Rigobon (2002), No Contagion, Only Interdependence: Measuring Stock Market Comovements, The Journal of Finance, Vol. LVII, No 5.

Forbes, K. and F. Warnock (2012), Capital Flow Waves: Surges Stops, Flight, and Retrenchment, Journal of International Economics Vol. 88, pp. 235-251.

Glasserman, P. and H. P. Young (2016), Contagion in Financial Networks, Journal of Economic Literature, Vol. 54, No 3, pp. 779-831.

Hossfeld, O. and M. Pramor, Global liquidity and exchange market pressure in emerging market economies, Deutsche Bundesbank Discussion Paper No 05/2018, April 2018.

International Monetary Fund (2020a), https://www.imf.org/en/Topics/imf-and-covid19/ Policy-Responses-to-COVID-19, 31 July 2020.

International Monetary Fund (2020b), Group of Twenty: Implementation of the G-20 Action Plan, July 2020.

International Monetary Fund (2016), Global Financial Stability Report, April 2016.

International Monetary Fund (2015), Assessing Reserve Adequacy – Specific Proposals, IMF Policy Paper, 121914.

International Monetary Fund (2011), Assessing Reserve Adequacy, IMF Policy Paper, 021411b.

Koepke, R. and S. Paetzold (2020), Capital Flow Data – A Guide for Empirical Analysis and Real-Time Tracking, IMF Working Paper, WP/20/171.

Koop, G., M. H. Pesaran and S.M. Potter (1996), Impulse response analysis in nonlinear multivariate models, Journal of Econometrics, Vol. 74, pp. 119-147.

Lütkepohl, H. (2005), New Introduction to Multiple Time Series Analysis, Springer Verlag Berlin Heidelberg.

Obstfeld, M. (2014), Trilemmas and Tradeoffs: Living with Financial Globalization, http://eml. berkeley.edu//~obstfeld/Trilemmas_last%20draft.pdf, 14 August 2020. Deutsche Bundesbank Monthly Report September 2020 73

Pagliari, M. S. and S. A. Hannan (2017), The Volatility of Capital Flows in Emerging Markets: Measures and Determinants, IMF WP/17/41.

Pesaran, H. H. and Y. Shin (1998), Generalized Impulse Response Analysis in Linear Multivariate Models, Economic Letters, Vol. 58, pp. 17-29.

Upper, C. (2011), Simulation methods to assess the danger of contagion in interbank markets, Journal of Financial Stability, Vol. 7, pp. 111-125.

Wooldridge, P. (2020), Implications of financial market development for financial stability in emerging market economies, https://www.bis.org/publ/othp32.htm, 14 August 2020.